Joule Thief

[Pirate88179]

@Pirate: There are too many variables to be able to tell if your proposed lightmeter method is valid or not. The integration time constant of the lightmeter is unknown for example. Better, I think, would be to use your oscilloscope or even the DMM to take a current reading (Vdrop across a known series small resistance) in combination with a voltage reading on the other channel; one can make a very good "guesstimation" of the necessary multiplication to get an average power value that way.

TK:

Thank you for the reply.  I would use one of my dmms but I do not know how to put the meter in series with the input now that I have wired the light in parallel. (with 3 separate inputs for all 3 AAA bats.) I suppose I could cut all of my connections and place the meter in series but, I was hoping to avoid all of that. Maybe I should just get another light and use another circuit board and test it prior to modding the battery inputs?  Hey, that way I can have a bunch of clip leads all over my bench and I can video that and claim overunity?  (Just kidding)

I appreciate the advice.

Thanks,

Bill

[MarkE]

Mark:

Wow, thanks, that is a lot of good info there.  I considered removing some of the leds but, that would not look too good in this package.

I will have to open the light up again to determine if the leds are in series, or parallel.

In those oval lights they are all in parallel.

I can not recall which way draws less Ma's at this moment.  In the reviews of this light on Amazon, there were complaints that it would only run for about 4 hours.  I figured with this modification, I could get maybe close to 12.

The one that I have would be very lucky to make 4 hours at any reasonable brightness.

I do not like the idea that I might be pushing 400 mA's through this board rated at only 200 mA's.  Nothing good can happen in the long run with numbers like that.  It would be very difficult for me to use a meter to check my actual amp draw, as everything is now soldered up in parallel.  I know TK has always said that you can not rely on your power supply to determine amp draw but, what if I took a reading with my light meter at a fixed location from the light, and then, used my power supply to get the same reading?  Would not the amp draw used by my digital power supply be at least close?  I can use my magnetic clip jumper wires to hook up my light and it would not be hard to do at all.

If you put a sizeable capacitor in parallel with the power supply the capacitor will even out the ripple and the power supply current reading will be reasonably accurate.

I just checked a laser yesterday that I was repairing/modifying for a friend of mine. ( I modified it to run on multiple AA batteries instead of those small button cells.  Mine is still working on the original AA bats. after playing with my cat every day for 6 months)  As I turned up the voltage on my power supply to 3 volts, I got a reading of 200 mA's which was the same on the laser I modified for myself.

Another alternative I was considering was to use a single 18650 Li-on battery at 3.7 volts and 4,000 mA hours and forget the expensive circuit board.  This is almost 4 times the mA rating of a AAA battery, which determines the longevity of light in the stock configuration in series.  That should give me somewhere around 12 hours of light as well...no?

Thanks for your help,

Bill

An 18650 will really yield about 2600 - 3000mAh.  There are many unscrupulous sellers who put arbitrary labels on cells and even sell used cells.  If it isn't Panasonic or Samsung then it's something of a crap shoot.


There are buck converters designed to work with Li-ion cells.  I have not worked with them.

[Pirate88179]

Well folks, MarkE was correct...the ncp1402 board in my harbor freight light has begun to flash in an erratic manner after only about 3+ hours on the 24 leds setting.

So, once again, it is back to the drawing board.

Thanks for the info Mark.

Bill

[Pirate88179]

SparkFun rates the chip as the data sheet does which is the voltage that the chip will start at.  This isn't much of a problem when running LEDs because the alkaline cell will still put out 0.8V under very light load until it is very depleted.  So, the NCP1402 bootstraps itself up to the 2.5V - 3V needed to get the LED going.  You can modify those oval Harbor Freight lights easily enough to disconnect half of the cheapo LEDs that they use.

The drop-out voltage will be sort of a soft thing.  What will happen is that the NCP1402 will run up against its maximum duty cycle and then the LED voltage will fall and the LED current will then rapidly collapse.  The NCP1402 is rated for 70% to 85% maximum duty cycle.  Assuming that you need at least 2.6V to run a white LED without a current limiting resistor, then you could limit as high as 0.78V or as low as 0.39V with 0.57V as a typical value.  If you need a current limiting resistor, you will start dimming out pretty close to 0.8V cell voltage.

Personally, if your goal is to get the most light times hours out of a particular light and battery combination, then I think you will be much better off addressing the LED choice and the power converter efficiency than worrying about extending the drop-out voltage below 0.8V or even 0.9V.  At 100mW, the additional energy that is available from a AA alkaline cell from 0.8V down to 0.5V is about 5%.  At 250mW load it is hard to measure.    With AAA's there is almost no run time extension even at 100mW.  If you can get them, you might want to play with PAM2803 or PAM2805 chips.  They are designed for LED flashlights and incorporate the current regulation while dropping a mere 95mV in the sense resistor.  They are rated to start at 0.9V.  They are hard to get in the USA.

Those oval lights that Harbor Freight sells have a 1.5 Ohm current limit resistor in series with the LEDs.  Driven by a bench supply at 4.5V (three fresh AAA's) they draw 440mA.  When the voltage drops to 3.6V they are down to 210mA.  At 3V they are down to 70mA a mere 1/6th of their initial brightness, and at 2.5V they are off.  440mA is very hard on a AAA cell, and a AAA cell can only go about 45 minutes at that kind of load.  Even if you leave their junk LEDs alone, you could greatly extend the run time at good brightness by getting rid of the dropping resistor and running the LEDs at say 200mA and 3.3V.  At 80% converter efficiency you would need 825mW from the cells.  At 825mW you should get about 2.5 hours run time.   That's more than triple what people can expect from those lights as they are. Part of which you get by running at about 200mA all the time instead of 440mA starting.  200mA output is more than the NCP1402 can handle but it is well within the capability of the PAM2803 and PAM2805.

ETA:  Here is an extrapolated plot of AA alkaline run time at constant power loads of 250mW and 100mW, and the unextrapolated AAA plot.

I found this online as a typical circuit to be used with the Pam2803.  Now all I have to do is find a source for the chips.

Thanks Mark,

Bill

[TinselKoala]

http://www.utsource.net/ic-datasheet/PAM2803-5062937.html

I've dealt with UTSource half a dozen times or so, and they are reliable and pretty fast. About 10 business days from China to my mailbox in Texas. Their pricing structure is kind of strange, but they do respond to email questions. Order some other stuff at the same time! They have obscure, obsolete, and hard to find semiconductors, as well as ordinary stuff.

I've taken apart one of the 24 LED oval lights from HF and rewired it so that it is 2-series-12-parallel instead of 24-parallel. This makes it work a little better as a load for my JTs, which seem to like pushing through the doubled fwd voltage.